Mobile game developer and publisher Pixel Federation entertains more than 140M players worldwide with popular mobile games, including ‘Diggy’s Adventure,’ ‘Trainstation 2,’ ‘Puzzle Adventure,’ and ‘Port City.’ Founded by four game enthusiasts in 2007, the Slovakia-based company specializes in creating mobile free-to-play community-based experiences, determining development priorities based on player feedback. Hampered by increasingly frequent issues with local infrastructure, Pixel Federation migrated its core functionality to Amazon Web Services (AWS) with support from AWS Partner Labyrinth Labs.

Puzzle Adventure

“When we release new games or features, our concurrent player count can multiply in seconds, and that demand is hard to support with on-prem infrastructure. We needed access to more compute and the ability to scale on demand, which led us to Labyrinth Labs and AWS,” said Branislav Bernat, CIO, Pixel Federation. “We evaluated a few cloud providers, and ultimately AWS provided the best experience and capacity; it was an easy decision.”

Pixel Federation initially took an approach based on open-source components to avoid relying on its existing on-premises provider. After running into challenges related to Apache Kafka and proxies, it enlisted cloud experts Labyrinth Labs for help. Upon evaluating Pixel Federation’s underlying technology, Labyrinth Labs proposed a plan for migrating the studio to Kubernetes on AWS for better stability, performance, and compute availability.

“Sometimes companies are reluctant to move to the cloud because it’s hard to compare costs directly. There are also intangibles associated with each, like on-premises infrastructure maintenance costs and unanticipated revenue loss during an infrastructure outage. But once you understand the variables and implement good cloud practices, it’s clearly the right path,” noted Adam Hamsik, Labyrinth Labs CEO and Co-founder.

After receiving the green light to move forward, Labyrinth Labs began Pixel Federation’s move to AWS with ‘Diggy’s Adventure,’ the developer’s most popular and profitable game. Also serving as Lead DevOps Engineer, Hamsik helped oversee the transition, along with Pixel Federation’s Lead Backend Software Architect for PHP projects, Martin Fris.

“The existing infrastructure was identified as a risk due to its frequent failures, so we prioritized our migration steps to minimize exposure to outages, which is why we moved the biggest title first,” explained Hamsik. “At the same time, we focused on building a strong cloud engineering culture, which provided the foundation for Pixel Federation’s AWS journey.”

Building with the basics

Pixel Federation started with a proof of concept (PoC) to modernize the legacy on-premises solution by moving it to the cloud. A simple lift and shift approach wasn’t an option, so the team opted to build a new solution on AWS, where they could migrate the data, servers, and workloads and install all components from scratch. The implementation was a huge effort that required significant time and resources.

To address issues with legacy infrastructure, the team evaluated multiple solutions to run the workload efficiently and reliably. After thorough discussions, Kubernetes was determined as the clear choice. Since the PoC predated the launch of Amazon Elastic Kubernetes Service (Amazon EKS), the team decided to build a Kubernetes cluster using open-source projects and launched ‘TrainStation 2’ there.

‘TrainStation 2’ became the first cloud-native game deployed by Pixel Federation with multiple iterations. The effort took almost a year, during which DevOps engineers worked on developing and deploying the target infrastructure and solutions for observability, automation, and other supporting functions. This time was also essential for the team to learn how to successfully run workloads in the cloud, in preparation for migrating ‘Diggy’s Adventure’, Pixel Federation’s biggest and most successful game to date, to the cloud.

“We learned a lot launching our first game on AWS. As with any deployment, testing was crucial. Once we had our software stack running smoothly, the game experience was great,” said Fris, Lead Backend Software Architect, Pixel Federation. “We’ve also been able to implement changes to lower our ongoing costs.”

Trainstation 2

Ongoing optimizations

Once the initial migration was complete, Pixel Federation and Labyrinth Labs began a refactoring period during which they looked at the pieces set in place and identified areas for improvement. The goal was to create a modular, sustainable solution that could be continually enhanced over time and was easier to maintain. In addition to leveraging its AWS services, the team decided to reinforce the existing setup with Amazon EKS. Prioritizing managed services enabled Pixel Federation to save operational costs and use auto-scaling with Amazon Elastic Compute Cloud (Amazon EC2) Spot Instances to better accommodate peaks in user demand, making the system more performant and cost-efficient.

“We also found that we could optimize cloud costs through good software architecture,” shared Hamsik. “We wanted better results – cheaper, faster, and more reliable, so we consulted with Martin (Chief Architect at Pixel Federation) on how best to achieve that. He suggested running games on a different PHP engine, which helped games run faster using fewer nodes. We now sit down with our programmers every month to look for new ways to reduce costs by refining processes.”

By 2021, all Pixel Federation workloads were running on AWS. “The migration to AWS was transparent from a player’s standpoint, but we definitely sleep better at night now. AWS is so much more reliable than our previous infrastructure; we can build and launch games without worrying about how to support the existing number of players,” said Bernat.

Only the company’s data warehouse remains on premises, though Pixel Federation is preparing to move that final piece to AWS as well. The team plans to take a similar approach to managing this data on the cloud with Kubernetes so that the same group can easily manage it. Shifting to AWS has also helped Pixel Federation gain better monitoring and observability capabilities. Hamsik explained, “The migration enabled us to better define and structure how we work with games. We now have the tools we need to better see and understand what’s happening on an infrastructure and application level.”

Diggy’s Adventure

Putting the final pieces in place

While Pixel Federation has been able to optimize operations with AWS, it has yet to unlock the potential of nearly 15 years of collected data. Its entire analytics stack is currently housed on-premises and comprises large amounts of analytics tools and solutions, making it complex and costly to maintain. By migrating its data warehouse to the cloud and going all-in on AWS, the company will be better positioned to gain data insights, as well as leverage artificial intelligence (AI) and machine learning (ML) services, like Amazon SageMaker, to further enhance the player experience.

In preparation for the move, the Labyrinth Labs team worked with AWS to become eligible for the Migration Acceleration Program (MAP), which is designed to help customers migrate to the cloud faster and offset costs.

“Now that we have all our infrastructure on AWS, we don’t ever plan to go back. We automate as much as we can so that we can focus more on important things, like growing our capabilities and making great games,” concluded Bernat. “We’re thankful for the support of AWS and Labyrinth Labs; the cloud journey is definitely easier when you have a great partner.”

For more information on using AWS for game development, visit: https://aws.amazon.com/games